Internals¶
We ran into three main problems developing this: Exceptions, callbacks and accessing socket methods. This is what this chapter is about.
Exceptions¶
We realized early that most of the exceptions would be raised by the I/O
functions of OpenSSL, so it felt natural to mimic OpenSSL’s error code system,
translating them into Python exceptions. This naturally gives us the exceptions
SSL.ZeroReturnError
, SSL.WantReadError
,
SSL.WantWriteError
, SSL.WantX509LookupError
and
SSL.SysCallError
.
For more information about this, see section SSL — An interface to the SSL-specific parts of OpenSSL.
Callbacks¶
Callbacks were more of a problem when pyOpenSSL was written in C. Having switched to being written in Python using cffi, callbacks are now straightforward. The problems that originally existed no longer do (if you are interested in the details you can find descriptions of those problems in the version control history for this document).
Accessing Socket Methods¶
We quickly saw the benefit of wrapping socket methods in the
SSL.Connection
class, for an easy transition into using SSL. The
problem here is that the socket
module lacks a C API, and all the
methods are declared static. One approach would be to have OpenSSL
as
a submodule to the socket
module, placing all the code in
socketmodule.c
, but this is obviously not a good solution, since you
might not want to import tonnes of extra stuff you’re not going to use when
importing the socket
module. The other approach is to somehow get a
pointer to the method to be called, either the C function, or a callable Python
object. This is not really a good solution either, since there’s a lot of
lookups involved.
The way it works is that you have to supply a socket
- like transport
object to the SSL.Connection
. The only requirement of this object is
that it has a fileno()
method that returns a file descriptor that’s
valid at the C level (i.e. you can use the system calls read and write). If you
want to use the connect()
or accept()
methods of the
SSL.Connection
object, the transport object has to supply such
methods too. Apart from them, any method lookups in the SSL.Connection
object that fail are passed on to the underlying transport object.
Future changes might be to allow Python-level transport objects, that instead
of having fileno()
methods, have read()
and write()
methods, so more advanced features of Python can be used. This would probably
entail some sort of OpenSSL BIOs, but converting Python strings back and
forth is expensive, so this shouldn’t be used unless necessary. Other nice
things would be to be able to pass in different transport objects for reading
and writing, but then the fileno()
method of SSL.Connection
becomes virtually useless. Also, should the method resolution be used on the
read-transport or the write-transport?